SQL (first and last record)

Question

Hi All,
WITH t AS
SELECT 100 region_id, 1001 Client_Id, 'client 1' Client_Name, 1 Status_id, 'progress' Status_name, SYSDATE + 1 Start_Date, SYSDATE + 1 End_Date FROM dual UNION ALL
SELECT 100 region_id, 1001 Client_Id, 'client 1' Client_Name, 2 Status_id, 'hold' Status_name ,SYSDATE + 1 Start_Date, SYSDATE + 1 End_Datet FROM dual UNION ALL
SELECT 100 region_id, 1001 Client_Id, 'client 1' Client_Name, 3 Status_id ,'reject' Status_name, SYSDATE + 1 Start_Date, SYSDATE + 1 End_Date FROM dual UNION ALL
SELECT 200 region_id, 1002 Client_Id, 'client 2' Client_Name, 1 Status_id, 'progress' Status_name, SYSDATE + 1 Start_Date, SYSDATE + 1 End_Date FROM dual UNION ALL
SELECT 200 region_id, 1002 Client_Id, 'client 2' Client_Name, 2 Status_id, 'hold' Status_name ,SYSDATE + 1 Start_Date, SYSDATE + 1 End_Datet FROM dual UNION ALL
SELECT 200 region_id, 1002 Client_Id, 'client 2' Client_Name, 3 Status_id ,'reject' Status_name, SYSDATE + 1 Start_Date, SYSDATE + 1 End_Date FROM dual UNION ALL
SELECT 200 region_id, 1002 Client_Id, 'client 2' Client_Name, 3 Status_id ,'progress' Status_name, SYSDATE + 1 Start_Date, SYSDATE + 1 End_Date FROM dual
SELECT * FROM t;Desired output:
REGION_ID     CLIENT_ID     CLIENT_NAME     STATUS_ID     STATUS_NAME     START_DATE                    END_DATE
100     1001     client 1                      1                     progress                     9/3/2011 12:26               9/3/2011 12:26
100     1001     client 1                      3                     reject                     9/3/2011 12:26                     9/3/2011 12:26
200     1002     client 2                      1                     progress                     9/3/2011 12:26                     9/3/2011 12:26thanks,

Answer 1

Hi,
Like Sven, I don't understand why you only want one output row for client_id=1002. Also, I don't understand what order determine "first" and "last". It's hard to guess because so many of the columns have the same value on every row.
My best guess is that status_id plays a part in deciding who is "first" and "last", and that you don't want to include ties (multiple rows with a equal claim to being "first" or "last") in the output.
WITH     got_analytics     AS
     SELECT     t.*     -- or whatever columns you want to display
     ,     ROW_NUMBER () OVER ( PARTITION BY  region_id
                         ,             client_id
                         ,             client_name
                         ORDER BY        end_date
                         ,             start_date
                         ,             status_id
                        )     AS a_num
     ,     ROW_NUMBER () OVER ( PARTITION BY  region_id
                         ,             client_id
                         ,             client_name
                         ORDER BY        end_date     DESC
                         ,             start_date     DESC
                         ,             status_id     DESC
                        )     AS d_num
     ,     COUNT (*)     OVER ( PARTITION BY  region_id
                         ,             client_id
                         ,             client_name
                         ,             end_date     -- NOTE: no ORDER BY
                         ,             start_date
                         ,             status_id
                        )     AS cnt
     FROM     t
--     WHERE     ...     -- if you need any filtering, this is where it goes
SELECT       region_id, client_id, client_name, status_id, start_date, end_date
FROM       got_analytics
WHERE       (     a_num     = 1
       OR     d_num     = 1
AND       cnt     = 1
ORDER BY  region_id, client_id, client_name, start_date, end_date, status_id
;If client_name and region_id depend on client_id, then you don't need to include them in the PARTITION BY clauses. Depending on your requirements, you may not want them in the PARTITION BY clauses even if they are independent of client_id. Then again, you may not want client_id in th PARTITION BY clauses. It all depends on what you're trying to do.